Changes in estimated glomerular filtration rate before and after the first visit for atrial fibrillation

Background Although the development of atrial fibrillation (AF) and the progression of chronic kidney disease are known to be interrelated, it remains unclear when and how renal function changes during the clinical course of AF. Methods This study retrospectively enrolled 131 patients who were able to collect data on estimated glomerular filtration rate (eGFR) at least five times during the 500 days before and 500 days after the first visit (baseline) of new-onset AF, respectively. To investigate the temporal relationship between the development of AF and the beginning of worsening renal function (WRF), a piecewise regression model was applied to the eGFR time series data. The time point at which the slopes of the two regression lines changed (inflection -point), the slope before and after the inflection-point (β1 and β2, respectively), and the difference in slope (Δβ) were estimated. The presence of WRF was defined as having the inflection-point at which both Δβ and β2 were < − 0.0083 mL/min/1.73 m2/day (corresponding to 3.03 mL/min/1.73 m2/year), and the corresponding the inflection-point was defined as the beginning of WRF. Results WRF was detected in 54 (41.2%) patients. The beginning of WRF were distributed at various times, but most frequently (23 of 54 patients) within 100 days before and after baseline. The presence of WRF was not associated with age, heart failure, or baseline eGFR, but was associated with positive β1 (odds ratio 30.5, 95% confidence interval 11.1–83.9, P < 0.01). Conclusion In nearly half of AF patients with WRF, the beginning of WRF was observed within a few months before or after the first visit for AF. Patients with a positive eGFR slope before the onset of AF are more likely to develop WRF after the onset of AF, suggesting that potential kidney damage may be underlying.


Introduction
Several investigators have reported that atrial fibrillation (AF) increases the risk of chronic kidney disease (CKD), and that CKD increases the risk of new-onset of AF [1][2][3].It has been recognized that AF and CKD are closely interrelated, with either being a possible cause and result of the other.However, almost all interest has focused on changes in renal function after the development of AF [4][5][6].We conducted a study to identify when and how renal function changes during the development of AF.

Study design and inclusion criteria
This was a single-center, retrospective study conducted at Kindai University Hospital from May 2011 to October 2018.Using clinical information from medical records, we identified and screened 546 outpatients with AF on anticoagulant therapy.Of all patients, those for whom data measuring renal function were obtained at least 5 times in each period over 500 days before and after the occurrence of AF were selected.The day of the first visit for the new-onset AF was defined as baseline (day 0).The enrollment criteria were set as follows: (i) documentation of newly diagnosed AF (presumed to have symptoms of AF within a few months prior to the time of baseline), (ii) estimated glomerular filtration rate (eGFR) of at least 15 ml/min/1.73m 2 at baseline, and (iii) there were less than three missing intervals of eGFR data (per 100-day interval) before and after baseline.A flowchart of the methods is shown in Fig. 1.

Estimation of renal function
Renal function was estimated from Japanese equation for eGFR as follows: eGFR (unit: mL/min/1.73m 2 ) = 194 × serum creatinine − 1.094 × age − 0.287 (× 0.739 if female sex) [7].The serum creatinine levels obtained at each visit were examined from the medical records.Because a fatal condition affects renal function, data within 60 days before death were excluded from the analysis.If blood sample data were not available on day 0, the eGFR on the day closest to day 0 (within 1 month) was substituted for the baseline data.

Inflection-point detection and slope estimation in the eGFR
A scatter plot of changes in eGFR for each individual patient was drawn.A piecewise regression model was fitted to the scatter plot to identify the inflection-point, which represents the point of greatest change in eGFR during the follow-up period, and to estimate the slope before and after the inflection-point.The reason for using this statistical model was to examine the temporal relationship between the maximum change in eGFR and the onset of AF.If the maximal change in eGFR occurred close to the time of the first visit for AF, then it could be assessed as directly related to the development of AF.Conversely, if the maximal change in eGFR occurs Fig. 1 Flowchart of methods for analyzing changes in eGFR before and after the first visit for AF.eGFR estimated glomerular filtration rate far from the time of the first visit for AF, the association between the maximal change and the development of AF is weak.Piecewise regression models are statistical models in which two lines are joined at an unknown inflection-point.The model and parameters of the piecewise regression analysis are as follows.
where Y is the value of the eGFR corresponding to X, X is a time variable, and the parameter I is the time variable at the inflection-point.The parameter ε was assumed to be an independent, additive error with a mean of zero, constant variance, and a finite absolute moment of some order > 2 [8,9].The inflection-points were statistically detected by using a maximum likelihood, where the maximum probability of a change in slope occurs.Estimated inflection-points included those with subtle, clinically insignificant slope changes.The estimated inflection points were utilized for classification based on slope criteria before and after the inflection-point, as described later in the next section.

Definition of worsening renal function
We divided the estimated statistical inflection-points into those with and without worsening renal function (WRF).WRF was defined as having an inflection-point with both Δβ and β2 of < − 0.0083 mL/min/1.73m 2 /day, and the corresponding inflection-point was defined as the beginning of WRF.This criterion can be expressed by the following equation: Cases in which renal function improves before the inflection point but declines thereafter, and the absolute values are approximately balanced, do not meet the criteria for WRF (i.e., are not defined as WRF).
This criterion is based on the fact that in some clinical trials, the rate of decline in eGFR is considered rapid if the rate exceeds 3 mL/min/1.73m 2 /year (0.0083 mL/ min/1.73m 2 /day in daily calculations) [10,11].

Patients
Initially, 222 patients were excluded for criteria (i).56 were excluded for criteria (ii) or (iii).124 patients were excluded for criteria (iv).14 patients were excluded for criteria (v).Finally, 131 patients were enrolled (Fig. 3).

Clinical characteristics
The characteristics of the 131 patients are shown in Table 1.

Anticoagulation therapy
Anticoagulation was initiated in 128 of 131 patients at our hospital.Of these, 125 patients were started on anticoagulation therapy on the day of their first visit due to AF, 2 patients were started 1 month after their first visit, and 1 patient was started 3 months after their first visit.There were 3 cases in which anticoagulation therapy was initiated by the primary care physician prior to referral to our hospital.In the referred cases, their exact date of anticoagulation initiation was not traceable, but we assume that most of them started anticoagulation therapy within 1-2 months of their referral to our hospital, based on the contents of the medical information form.112 patients were started on anticoagulant therapy with direct oral anticoagulants (apixaban in 14 patients, rivaroxaban in 23 patients, edoxaban in 16 patients and dabigatran in 59 patients) at the first visit for AF.Changes in direct oral anticoagulants during the study period were not followed.19 patients received warfarin at the first visit and were switched to a direct oral anticoagulant after the second visit.Of these, one patient was switched to apixaban, one to rivaroxaban, three to edoxaban and 14 to dabigatran.

Factors associated with WRF
Positive β1 was significantly associated with the presence of WRF (odds ratio 30.5, 95% confidence interval 11.1-83.9,P < 0.01).Even after adjustment in a model including all clinical variables investigated at the first visit for AF, a positive β1 was associated with the presence of WRF (odds ratio 51.5, 95% confidence interval 14.6-181.9,P < 0.01).(Table 2).

Discussion
The main finding of this study is that in almost half of the AF patients who developed WRF, WRF begins in the 100 days before and after the first visit for AF.Although the association between AF and WRF has been shown in several papers, there are no data that use time series eGFR data to estimate the time point at which WRF begins and examine the temporal association with the onset of AF.Furthermore, our findings indicated that an increase in eGFR slope prior to WRF was associated with a subsequent rapid decline in eGFR slope.This finding suggests that the presence of potential renal impairment is a condition that predisposes to the development of cardiorenal syndrome in patients with AF.

Mechanism by which AF precedes WRF
Our findings indicated that in almost half of the AF patients who developed WRF, WRF began in early onset of AF.Renal hypoxia can be a powerful explanation for the effect of AF on renal function.In an animal model, Freidman et al. reported that AF produced a decrease in cardiac output, leading to decreased blood flow in the renal cortex [18].Pruijm et al. demonstrated that reduced renal cortical oxygenation predicted subsequent eGFR decline by the blood oxygenation level-dependent magnetic resonance imaging [19].It is known that in response to renal hypoperfusion, the tubuloglomerular feedback will in turn activate to reduce GFR, resulting in a decrease the metabolic demand and the elimination of oxygen deprivation [20].And this feedback system breaks down due to continued renal hypoxia, leading to irreversible kidney damage [21,22].WRF might be more commonly observed in persistent AF than paroxysmal AF, since persistent AF is exposed to hemodynamics changes due to AF longer than paroxysmal AF.However, a previous report indicated that hospitalization due to heart failure is more common in patients with paroxysmal AF than persistent AF [23].In patients with paroxysmal AF, acute and repeated changes in heart rate and blood pressure during each AF episode may provide adverse effect on renal function or may not provide enough time to adapt AF hemodynamics.Therefore, we consider that the hemodynamic changes resulting in WRF were not always greater in persistent AF than in paroxysmal AF.GFR is essentially regulated by autoregulation of renal blood flow, however, Hill et al. demonstrated the morphological correlates of loss of autoregulation in the aging kidney of humans [24].Our findings indicate that AF patients with positive β1 are more susceptible to subsequent adverse effects on renal function, and suggest that positive β1 may represent nephrons overload.Excessive elevation of eGFR is a marker of early renal damage called "hyperfiltration, " which is known to lead to exhaustion of the remaining nephrons and CKD progression [25].Although eGFR slope was not included as known criteria for hyperfiltration, we believe that positive β1 may indicate hyperfiltration.This is because the kidney undergoes a number of morphological changes with age, as well as a natural gradual decline in clearance function [26,27].Therefore, we speculate that a positive β1 against agerelated eGFR decline may indicate nephrons overload, and nephrons overload preceding the development of AF may cause exhaustion of the nephrons, resulting in an early eGFR decline after the development of AF.We speculated that the elderly AF patients included those with loss of autoregulation of renal perfusion and glomerular overload, and that the nephrons in these patients were susceptible to the hemodynamic effects of AF, resulting in the onset of WRF early after the development of AF.

Clinical implications
Kidney management should be initiated at the time of first visit for AF, even in patients with normal renal function.For proper management of direct oral anticoagulation, renal function should be reconfirmed within the first few months after the first visit for AF.

Limitations
This study has the following limitations.First, because this study was a retrospective, single-center study, the frequency and the interval of available eGFR data were different for each patient.Second, we could not investigate data on proteinuria, which is a known risk factor for CKD progression [28].Thirdly, we could not investigate the type, dosage, and duration of diuretic medication.
In addition, changes in drug therapy including diuretics and ACE-Is/ARB may affect renal function, but this study does not present follow-up data on drug cessation or addition.Fourth, we could not analyze the effect of AF treatment on renal function.The implications of the improvement in eGFR slope are not clear in this study.
Treatment of AF and its comorbidities, such as heart failure, may have had a positive impact on renal function.Finally, because it was difficult to determine the date of the first AF episode, the date of the first AF diagnosis was set as the baseline.Patients who had subclinical AF more than a few months before the first visit for AF due to mild subjective symptoms may have been included in this study.

Conclusion
In nearly half of AF patients with WRF, the beginning of WRF was observed within a few months before or after the first visit for AF.Patients with a positive eGFR slope before the onset of AF are more likely to develop WRF after the onset of AF, suggesting that potential kidney damage may be underlying.

Fig. 2
Fig. 2 Scatterplots of the change in the eGFR in two representative cases.Scatterplot of a patient with the point of WRF (a).Scatterplot of a patient without the point of WRF (b).C the inflection-point, β1 slope of the regression line before the inflection-point (units: mL/min/1.73m 2 /day), β2 slope of the regression line after the inflection-point (units: mL/min/1.73m 2 /day), Δβ β2 -β1(units: mL/min/1.73m 2 /day)

Fig. 5 Fig. 4
Fig. 5 Frequency distribution in the inflection-point with WRF.Box plot indicates median of day − 2, 1st quantile of day − 139, and 3rd quantile of day 100.Green curve indicates the normal distribution fit.WRF worsening renal function

Fig. 6
Fig. 6 Bivariate relationships between the inflection-point with WRF and β1.WRF worsening renal function, β1 slope of the regression line before the inflection-point (units: mL/min/1.73m 2 /day)

Table 2
Multivariate analysis of clinical variables and WRF Data are mean ± standard deviation or n (%) Abbreviations: CHF, chronic heart failure; LAD, left atrial dimension; LVEF, left ventricular ejection fraction; ACE-Is, angiotensin-converting enzyme inhibitors; ARBs, angiotensin-receptor blockers